Use of fuzzy similarity index for epileptic seizure prediction

A fuzzy similarity index is proposed to indicate the preictal state with EEG signal. First, during the process of calculating the correlation integral, a Heavyside function is replaced by a Gaussian function, which eliminates the effect of the crisp boundary of the Heavyside since the Gaussian function´s boundary is not sharp. Second, using a real EEG to compare the fuzzy similarity index and dynamical similarity index, it is found that the fuzzy similarity index is insensitive to the selection of the radius value and the EEG signal length. Finally, the fuzzy similarity index is applied to indicate the preictal state of a rat with EEG signal. The result shows that the performance of fuzzy similarity index of predicting epileptic seizure is better than that of dynamical similarity index.